At many banks, AI projects have been primarily IT-driven. A better model involves IT leaders collaborating with colleagues within the business to find the best path forward. These collaborations have resulted in some of the most impressive AI deployments the market has seen to date. Some banks are using digital twins in their treasury services to simulate cash flow scenarios and optimize intraday liquidity. Some companies are automating the entire order-to-cash lifecycle of commercial banking, with AI agents handling credit risk assessment, contract onboarding, and invoice generation and collections.
4. Human-centered
Technology alone won’t deliver the AI revenue banks are seeking. A skilled AI team, human-machine collaboration, and change management are key to success. Training and upskilling efforts will teach workers how to use AI to improve their jobs, especially when tied to comprehensive workforce transformation programs. In addition, the joint EY/MIT report3 We found that more than three-quarters of executives now view agent AI as a colleague. This is a fundamental shift in how workflows and governance models are designed.
Consider how junior analysts review initial AI output when creating a trading book. Meanwhile, senior underwriters will verify basic credit decisions made by AI and use AI-driven analytics to spot anomalies in loan portfolios. Commercial bankers use CoPilot to consolidate customer interaction history, assess overall risk posture, and flag opportunities to increase engagement with individual customers. These human-involved processes not only increase productivity, but also help reduce hallucinations and build trust in AI results.
As AI becomes more prominent across operations, banks will also create entirely new roles (such as prompt engineers, AI workflow designers, bot whisperers, and agent wranglers). These steps shorten the path to value and reduce risk along the way. Clear communication, visible leadership, and other proven organizational change management practices can instill an AI-positive mindset within your culture. “The biggest barrier to scaling AI is not the algorithms, but change management,” says Gupta. “Training teams, rebuilding processes, and putting in place the right governance often takes twice as much effort as building the model itself.”
Research from EY and others provides reason for optimism. The EY US Agentic AI Workforce Survey (via EY.com US) found that 84% of desk-based employees are enthusiastic about working with an AI agent. However, 56% are concerned about job security, illustrating the contradictions that bank leaders must address.
5. Futuristic and long-lasting design
While there is an urgent need for banks to accelerate their AI journey, leaders must recognize the length of the game and consider how AI (and agent AI) will reshape the future of banking. This means looking to the horizon and being prepared to pivot as technology advances and new capabilities create new possibilities. It’s certainly not too early to start thinking about how AI will interact with digital assets, tokenization, quantum computing, and other next-generation technologies. Once again, today’s breakthrough innovations will be tomorrow’s bets when it comes to AI.
Long-term planning must be reflected in the design of AI technology infrastructure, with an emphasis on modularity, reusability, and scalability. As vendors (from startups to SaaS platforms) mature their AI capabilities, baseline capabilities may be built into their products, eliminating the need for banks to develop their own. While cloud environments are essential, regulated banks are also considering other options for their most sensitive use cases (e.g., using their own large-scale language models in on-premises technical environments). Such hybrid approaches are likely to become more common in the future as banks seek to balance risk and return and enable innovation within compliant processes. Banks must also continually evaluate their vendor relationships and be prepared to change course as needs and priorities evolve and new types of solutions emerge.
